Method of License Plate Recognition Based on Character Reconstruction

A method of license plate recognition based on character reconstruction was presented.The character of license plate had more noise and hackle after binarization,and the character would be close to standard mode by character reconstruction,which was convenient for character recognition.In reconstruction,by the capability of associative memory,a set of normal modes was accumulated in the discrete Hopfield network.Firstly,the set of normal modes was changed into normal orthogonal vectors by the orthogonalization of Schimidt.Then,the Hebb rule was used to learn the matrix of weighted value,and the capability of associative memory was improved.In the end,a batch of samples was tested.The result shows that character reconstruction can noticeably improve the ratio of license plate recognition.